Information processing method and information processing apparatus

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to an information processing method and an information processing apparatus that are suitable when used for combination optimization problems such as traveling salesman problem, for example.

2. Description of the Related Art

Here is taken a problem in which the number N of its factors can be defined, as a problem for which a calculation should be executed. The problem for finding a route shortening the length of each line connecting points among routes once passing all of a plurality of points given on a two-dimensional plane is typical one of such problems, and the number of given points is represented by N. Considering these points as positions of cities, the problem of finding the shortest route is sometimes called a traveling salesman problem.

In general, in case that the calculation time required for executing a solution by naive algorithm increases by power of N, the problem is called a polynominal problem. On the other hand, in case that the calculation time increases exponentially of N, the increase of the calculation time with the increase of N is remarkable, and it is difficult to execute numerical calculation using a conventional computer. The class containing this problem is called NP (nondeterministic polynominal) complete problem. The complete NP problem has the possibility of becoming a polynominal problem if a dexterous algorithm is found, and vigorous researches are being made. It is important that, once any dexterous algorithm is found, all of the problems known as NP complete problems are shown to become polynominal problems.

The subject of the widest researches as an NP complete problem is the traveling salesman problem mentioned above. An example thereof is schematically shown in FIG. <b>1</b>. In FIG. 1, white points represent positions of cities, and the set of lines each connecting two cities is the shortest route. To solve this kind of problem, various approximate methods have heretofore been proposed, such as those using neural networks or spin glass.

Recently, a new approximate method using the concept of renormalization group was found by Y. Usami and Y. Kano regarding the traveling salesman problem and has been remarked (Phys. Rev. Lett. 75, 1683 (1995)). The renormalization group is a method that has performed its power in analyses like phase transition, and it is a basic concept of modern physics. FIG. 2 schematically shows this method. In FIG. 2, black points represent positions of cities. This method divides a given distribution of cities into some regions called frames, and executes calculation for each frame to obtain an approximate solution. In the example of FIG. 2, a distribution is divided into four frames.

The approximate method using renormalization group is certainly excellent in capability of obtaining a solution more quickly than conventional methods. However, since it executes calculation for fixed frames, approximation accuracy cannot be increased so much, and it is difficult to realize an apparatus for execution and processing by a physical system as an exclusive device.

OBJECT AND SUMMARY OF THE INVENTION

It is therefore an object of the invention to provide an information processing method and an information processing apparatus capable of quickly obtaining excellent approximate solutions of a combinatorial optimization problem such as traveling salesman problem and enabling realization of the processing apparatus as massively parallel exclusive devices.

Toward attainment of the above-mentioned object, the Inventor developed the concept of the above-indicated renormalization group, then contrived the use of a new renormalization transformation using movable frames instead of fixed frames and found the possibility of using this technique to obtain approximate solutions of combinatorial optimization problems such as traveling salesman problem with a high accuracy and at a high speed, which is just the present invention.

According to the first aspect of the invention, there is provided an information processing method comprising: preparing an information carrier corresponding to a distribution of a plurality of points given on an n-dimensional space (where n is an integer not smaller than 2); and using time development and time reversal of the information carrier to process the information.

According to the second aspect of the invention, there is provided an information processing apparatus configured to prepare an information carrier corresponding to the distribution of a plurality of points given on an n-dimensional space (where n is an integer not smaller than 2), and to use time development and time reversal of the information carrier for processing the information.

In the present invention, global nature of information given on an n-dimensional space, in particular, is detected by preparing an information carrier corresponding to a distribution of a plurality of given points and using the time development and time reversal of the information carrier.

In the present invention, the information carrier may be the density of particles (such as atoms or molecules) corresponding to a distribution of a plurality of given points, and its diffusion process may be used as a time change (time development). Alternatively, the information carrier may be an optical intensity corresponding to a distribution of a plurality of given points, and its defocusing process may be used as the time change (time development).

In a typical example of the present invention, for a problem of finding a route minimizing the total length of lines connecting points from routes once passing all points given on the n-dimensional plane (traveling salesman problem), an information carrier corresponding to a distribution of the given points is prepared, and time reversal of the information carrier is used.

Information processing according to the invention is such that all of its processes can be executed with a computer (or IC), and so may be done. However, physical processes, such as defocusing process and diffusion process, can be executed by using an existent physical system.

According to the invention having the above-summarized structure, good approximate solutions of combinatorial optimization problems such as traveling salesman problem can be calculated at a high speed by thinning movable frames of renormalization transformation while moving them to meet with a given distribution of a plurality of points on an n-dimensional space, such as distribution of cities. Since the renormalization transformation can be executed by using time reversal of a simple physical phenomenon such as physical process like diffusion process or defocusing process, it can be realized in form of massively parallel exclusive devices.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram that explains a traveling salesman problem;

FIG. 2 is a schematic diagram that explains a method of solving a traveling salesman problem by renormalization transformation using fixed frames;

FIG. 3 is a schematic diagram that shows a solution of a traveling salesman problem using movable-frames renormalization transformation;

FIG. 4 is a schematic diagram that shows a solution of a traveling salesman problem using movable-frames renormalization transformation;

FIG. 5 is a schematic diagram that shows a solution of a traveling salesman problem using movable-frames renormalization transformation;

FIG. 6 is a flow chart that shows procedures of a method of solving a traveling salesman problem using movable-frame renormalization transformation;

FIG. 7 is a schematic diagram that shows Γ(x, 0; R) upon a change of R;

FIG. 8 is a schematic diagram that shows φ(x, 0; R) upon a change of t;

FIG. 9 is a schematic diagram for explaining a method of solving a traveling salesman problem using movable-frame renormalization transformation;

FIG. 10 is a schematic diagram for explaining a method of solving a traveling salesman problem using movable-frame renormalization transformation;

FIG. 11 is a schematic diagram for explaining a method of solving a traveling salesman problem using movable-frame renormalization transformation;

FIG. 12 is a schematic diagram for explaining a method of solving a traveling salesman problem using movable-frame renormalization transformation;

FIG. 13 is a schematic diagram that shows a method of solving a traveling salesman problem according to the first embodiment of the invention; and

FIG. 14 is a schematic diagram that shows a method of solving a traveling salesman problem according to the second embodiment of the invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Prior to explaining embodiments of the invention, here is made a review about moving-frame renormalization transformation. This method generates frames in accordance with distribution of given cities, and deals with the problem, recognizing responses upon changes in size of the frames. As a possible method, discs are considered, which each have the radius R from each of cities C_i(i=1, 2, . . . , N) given on a two-dimensional plane (x-y plane), and a function taking 1 inside and 0 outside is used. A two-dimensional step function h (x, y; R) is defined as <maths><math><mtable><mtr><mtd><mrow><mrow><mi>h</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><mo>{</mo><mtable><mtr><mtd><mn>1</mn></mtd><mtd><mrow><mrow><mi>when</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><msqrt><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup></mrow></msqrt></mrow><mo>&lt;</mo><mi>R</mi></mrow></mtd></mtr><mtr><mtd><mn>0</mn></mtd><mtd><mrow><mrow><mi>when</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><msqrt><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup></mrow></msqrt></mrow><mo>&gt;</mo><mi>R</mi></mrow></mtd></mtr></mtable></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>1</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00001" file="US06581002-20030617-M00001.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00001" attachment-type="nb" file="US06581002-20030617-M00001.NB" /></attachments></maths>

When the position of the i-th city is expressed by (X_i, Y_i) (i=1,2, . . . , N), <maths><math><mtable><mtr><mtd><mrow><mrow><mi>G</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>N</mi></munderover><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mrow><mi>h</mi><mo></mo><mrow><mo>(</mo><mrow><mrow><mi>x</mi><mo>-</mo><msub><mi>X</mi><mi>i</mi></msub></mrow><mo>,</mo><mrow><mrow><mi>y</mi><mo>-</mo><msub><mi>Y</mi><mi>i</mi></msub></mrow><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>2</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00002" file="US06581002-20030617-M00002.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00002" attachment-type="nb" file="US06581002-20030617-M00002.NB" /></attachments></maths>

is useful. However, this function is not differentiable, introduction of a differentiable function is convenient. For example, <maths><math><mtable><mtr><mtd><mrow><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><mn>1</mn><mo>+</mo><mrow><mi>exp</mi><mo></mo><mrow><mo>[</mo><mrow><mi>β</mi><mo></mo><mrow><mo>(</mo><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup><mo>-</mo><msup><mi>R</mi><mn>2</mn></msup></mrow><mo>)</mo></mrow></mrow><mo>]</mo></mrow></mrow></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>3</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00003" file="US06581002-20030617-M00003.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00003" attachment-type="nb" file="US06581002-20030617-M00003.NB" /></attachments></maths>

is a candidate. Equation (4) made by using Equation (3), namely, <maths><math><mtable><mtr><mtd><mrow><mrow><mi>Γ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>N</mi></munderover><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mrow><mi>x</mi><mo>-</mo><msub><mi>X</mi><mi>i</mi></msub></mrow><mo>,</mo><mrow><mrow><mi>y</mi><mo>-</mo><msub><mi>Y</mi><mi>i</mi></msub></mrow><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>4</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00004" file="US06581002-20030617-M00004.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00004" attachment-type="nb" file="US06581002-20030617-M00004.NB" /></attachments></maths>

is differentiable, and gradually approaches (x, y; R) in , β→∞. Γ(x, y; R) is a function that expresses initial city distribution with the accuracy R.

Here are remarked points (P_j, Q_j) where r(x, y; R) has the maximum value. The number M(R) of points having the maximum value depends upon R. In R<b>6</b>∞, all of the cities are averaged, and it results in M(R→∞)=1. On the other hand, when R decreases, Γ(x, y; R) results in having values merely near cities, it results in M(R→0)=N. Let both these extremities be connected. As R decreases from ∞, a point having the maximum value will separate at a certain level, and it will result in M(R)=2. This occurs when distribution of cities is roughly divided into two groups and the distance of the groups becomes R. If R is further decreased, a point in one of the groups, which gives the maximum value, will separate, and it will result in M(R)=3. This occurs when distribution of cities is roughly divided into three groups and the distance among the groups becomes R. In this manner, as R changes from ∞ to 0, points giving the maximum value of Γ(x, y; R) progressively separate, and this finally terminates in the initial city distribution. Moving-frame renormalization transformation herein means that discs having the radius of R about points giving the maximum value of Γ(x, y; R) are considered as frames (moving frames) and they progressively move with changes of R (renormalization transformation). While executing the transformation, a branching diagram of points giving the maximum value is formed. By previously solving the traveling salesman problem for M(R)=k, the traveling salesman problem of M(R)=k+1 can be solved with reference to the solution for M(R)=k when branching occurs (M(R)=k→M(R)=k+1). That is, by solving a local problem of a portion where branching occurs, another solution is obtained by adding the remainder portion.

It is remarkable that this process can be realized by using defocusing in an optical system. If defocusing of an image of city distribution is progressed through a lens system, optical luminance profile obtained must be similar to Γ(x, y; R). Therefore, by taking an image of a city distribution with a video camera and oppositely rotating a movie made by defocusing, branching diagrams of points having the maximum luminance can be made progressively. Once such a branching diagram is obtained, it is the solution of the traveling salesman problem. Further, if it is recalled that, by progressing β→∞ with Γ(x, y; R), the result approaches the Gaussian distribution, it can be apparently realized with a diffusion system. For example, here is made a review about solutions of a two-dimensional diffusion process shown below. <maths><math><mtable><mtr><mtd><mrow><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><mfrac><mn>1</mn><mrow><mn>4</mn><mo></mo><mi>D</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mi>t</mi></mrow></mfrac><mo></mo><mrow><mi>exp</mi><mo></mo><mrow><mo>(</mo><mrow><mo>-</mo><mfrac><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup></mrow><mrow><mn>4</mn><mo></mo><mi>D</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mi>t</mi></mrow></mfrac></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>5</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00005" file="US06581002-20030617-M00005.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00005" attachment-type="nb" file="US06581002-20030617-M00005.NB" /></attachments></maths>

Superposition of the solutions: <maths><math><mtable><mtr><mtd><mrow><mrow><mi>Φ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>N</mi></munderover><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mrow><mi>x</mi><mo>-</mo><msub><mi>X</mi><mi>i</mi></msub></mrow><mo>,</mo><mrow><mrow><mi>y</mi><mo>-</mo><msub><mi>Y</mi><mi>i</mi></msub></mrow><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>6</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00006" file="US06581002-20030617-M00006.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00006" attachment-type="nb" file="US06581002-20030617-M00006.NB" /></attachments></maths>

is also a solution of the diffusion equation: <maths><math><mtable><mtr><mtd><mrow><mrow><mfrac><mo>∂</mo><mrow><mo>∂</mo><mi>t</mi></mrow></mfrac><mo></mo><mi>Φ</mi></mrow><mo>=</mo><mrow><mrow><mi>D</mi><mo></mo><mrow><mo>(</mo><mrow><mfrac><msup><mo>∂</mo><mn>2</mn></msup><mrow><mo>∂</mo><msup><mi>x</mi><mn>2</mn></msup></mrow></mfrac><mo>+</mo><mfrac><msup><mo>∂</mo><mn>2</mn></msup><mrow><mo>∂</mo><msup><mi>y</mi><mn>2</mn></msup></mrow></mfrac></mrow><mo>)</mo></mrow></mrow><mo></mo><mi>Φ</mi></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>7</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00007" file="US06581002-20030617-M00007.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00007" attachment-type="nb" file="US06581002-20030617-M00007.NB" /></attachments></maths>

Assuming t=R², the same discussion can be developed. Therefore, also by using time reversal of the diffusion process, moving-frame renormalization transformation can be executed to solve the traveling salesman problem. FIGS. 3, <b>4</b> and <b>5</b> show examples of solution using a diffusion process when N=20, 100 and 200, respectively.

A series of processing explained above can be shown by a flow chart of FIG. 6 as using a diffusion process.

Since the moving-frame renormalization transformation can be executed by using reverse transformation of a simple physical change such as defocusing or diffusion process, it can be realized by devices using a physical system. Then, by obtaining a solution of a traveling salesman problem, global nature of a given initial distribution can be grasped and detected.

A change with time of a distribution function of an information carrier is shown, taking the simplest case with two points as initial information. Let (X₁, Y₁)=(−1, 0) and (X₂, Y₂)=(1, 0) employed as those two points.

First shown is a change of an optical intensity distribution function in case of using defocusing in an optical system. This corresponds to the use of the step function: <maths><math><mtable><mtr><mtd><mrow><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mfrac><mn>1</mn><mrow><mn>1</mn><mo>+</mo><mrow><mi>exp</mi><mo></mo><mrow><mo>[</mo><mrow><mi>β</mi><mo></mo><mrow><mo>(</mo><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup><mo>-</mo><msup><mi>R</mi><mn>2</mn></msup></mrow><mo>)</mo></mrow></mrow><mo>]</mo></mrow></mrow></mrow></mfrac></mrow></mtd><mtd><mrow><mo>(</mo><mn>8</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00008" file="US06581002-20030617-M00008.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00008" attachment-type="nb" file="US06581002-20030617-M00008.NB" /></attachments></maths>

As an example, FIG. 7 shows results of <maths><math><mtable><mtr><mtd><mrow><mrow><mi>Γ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mn>0</mn><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mn>2</mn></munderover><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mrow><mi>x</mi><mo>-</mo><msub><mi>X</mi><mi>i</mi></msub></mrow><mo>,</mo><mrow><mrow><mi>y</mi><mo>-</mo><msub><mi>Y</mi><mi>i</mi></msub></mrow><mo>;</mo><mi>R</mi></mrow></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>9</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00009" file="US06581002-20030617-M00009.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00009" attachment-type="nb" file="US06581002-20030617-M00009.NB" /></attachments></maths>

in which β=20, and R is 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8 and 2.0, respectively. In FIG. 7, curves corresponding the respective values of R are shifted vertically for an easier view. It is known from FIG. 7 that the number points having the maximum value changes from two points to one point along with the increase of R.

Next shown is a change of a grain density distribution function in case of using a diffusion system. Here is used D=¼ in: <maths><math><mtable><mtr><mtd><mrow><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mi>y</mi><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><mfrac><mn>1</mn><mrow><mn>4</mn><mo></mo><mi>D</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mi>t</mi></mrow></mfrac><mo></mo><mrow><mi>exp</mi><mo></mo><mrow><mo>(</mo><mrow><mo>-</mo><mfrac><mrow><msup><mi>x</mi><mn>2</mn></msup><mo>+</mo><msup><mi>y</mi><mn>2</mn></msup></mrow><mrow><mn>4</mn><mo></mo><mi>D</mi><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mi>t</mi></mrow></mfrac></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>10</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00010" file="US06581002-20030617-M00010.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00010" attachment-type="nb" file="US06581002-20030617-M00010.NB" /></attachments></maths>

and points given by <maths><math><mtable><mtr><mtd><mrow><mrow><mi>Φ</mi><mo></mo><mrow><mo>(</mo><mrow><mi>x</mi><mo>,</mo><mrow><mn>0</mn><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow><mo>=</mo><mrow><munderover><mo>∑</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mn>2</mn></munderover><mo></mo><mstyle><mtext> </mtext></mstyle><mo></mo><mrow><mi>ϕ</mi><mo></mo><mrow><mo>(</mo><mrow><mrow><mi>x</mi><mo>-</mo><msub><mi>X</mi><mi>i</mi></msub></mrow><mo>,</mo><mrow><mrow><mi>y</mi><mo>-</mo><msub><mi>Y</mi><mi>i</mi></msub></mrow><mo>;</mo><mi>t</mi></mrow></mrow><mo>)</mo></mrow></mrow></mrow></mrow></mtd><mtd><mrow><mo>(</mo><mn>11</mn><mo>)</mo></mrow></mtd></mtr></mtable></math><img id="EMI-M00011" file="US06581002-20030617-M00011.TIF" img-content="math" img-format="tif" alt="embedded image" /><attachments><attachment idref="MATHEMATICA-00011" attachment-type="nb" file="US06581002-20030617-M00011.NB" /></attachments></maths>

by changing t as 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8 and 2.0, respectively, are plotted in FIG. <b>8</b>. It is known from FIG. 8 that the number of points having the maximum value changes from two points to one point here again along with the increase of t.

For easier understanding of the method of movable-frame renormalization group, a simple example is explained with reference to drawings. Assume that black points in FIG. 9 represent initial positions of cities. FIG. 10 shows an example in which a traveling salesman problem with five frames has been solved, with a turn line. In FIG. 11, movable frames are smaller. Therefore, although the cities are covered by seven frames, it can be regarded that central one and bottom-right one of five frames are separated. Therefore, using the solution of the traveling salesman problem with five frames, a traveling salesman problem with seven frames can be easily solved, and its solution is the turn line of FIG. <b>11</b>. FIG. 12 shows an example in which movable frames have been made smaller into 13 frames. FIG. 12 shows the solution of a traveling salesman problem with 13 frames by using the solution of FIG. <b>11</b>. In this manner, traveling salesman problems are sequentially solved while changing the size of movable frames. Once the process is executed until each frame finally contains only one city, the solution of the target problem can be obtained.

FIG. 13 shows an information processing method according to the first embodiment of the invention. The first embodiment uses a defocusing process for changes with time.

As shown in FIG. 13, in the first embodiment, a lens <b>12</b> is disposed such that a screen <b>11</b> indicating an initial city distribution is mirrored, and CCD <b>13</b> as an imaging device is positioned near the center of the lens <b>12</b>. FIG. 13 shows CCD <b>13</b> as being provided in a location distant by d from the focalized point. Output terminal of CCD <b>13</b> is connected to a computer <b>14</b>.

In the first embodiment, it is remarked that the image of the city distribution mirrored on CCD <b>13</b> can be defocused by progressively changing d. Then, while increasing d, starting from d=0, images of city distributions on the screen <b>11</b> are photographed with CCD <b>13</b>, and the traveling salesman problem of the given city distribution can be progressively solved on the basis of the image data.

FIG. 14 shows an information processing method according to the second embodiment of the invention. The second embodiment uses a diffusion process for changes with time.

As shown in FIG. 14, in the second embodiment, a liquid <b>22</b> is contained in a transparent container <b>21</b> open at the top, and a plurality of medicine droppers <b>23</b> are located above the liquid <b>22</b>. These medicine droppers <b>23</b> are arranged in a distribution coincident with the initial city distribution. Below the container <b>21</b>, a CCD camera <b>24</b>, for example, is located to take a picture of the liquid <b>22</b> inside the container <b>21</b> from below. The CCD camera <b>24</b> is connected to a computer <b>25</b> for processing image data.

In the second embodiment, when a colored liquid, such as a dye, is dropped into the liquid <b>22</b> like water, for example, the dye spreads in the liquid <b>22</b>. This process is recorded from time to time with the CCD camera <b>24</b>, and the data obtained are used to progressively solve the traveling salesman problem of the given city distribution.

Having described specific preferred embodiments of the present invention with reference to the accompanying drawings, it is to be understood that the invention is not limited to those precise embodiments, and that various changes and modifications may be effected therein by one skilled in the art without departing from the scope or the spirit of the invention as defined in the appended claims.

As described above, according to the invention, since an information carrier corresponding to the distribution of a plurality of points given on an n-dimensional space is prepared and information is processed by using time development and time reversal of the information carrier, good approximate solutions of a combinatorial optimization problem like a traveling salesman problem can be obtained quickly, and its processing apparatus can be realized in form of massively parallel exclusive devices.